Li Tan and Axel Krings. In Proceedings of the The 11th International Conference on Computational Science and Its Applications. Santander, Spain. In the Lecture Notes in Computer Science, Springer-Verlag. June 20-23, 2011.
This paper discusses the models and performance analysis for an adaptive software architecture, which supports multiple levels of fault detection, masking, and recovery through reconfiguration. The architecture starts with a formal requirement model defining multiple levels of functional capability and information assurance. The architecture includes a multi-layer design to implement the requirements using N-variant techniques. It also integrates a reconfiguration mechanism that uses lower layers to monitor higher layers, and if a fault is detected, it reconfigures a system to maintain essential services. We first provide a general reliability model (based on generalized stochastic Petri nets) for such a system with cross-monitoring for reconfiguration. Next, we define a probabilistic automaton-based model for behavioral modeling of the system. This model is especially suitable for modeling security problems induced by value faults. Whereas the Petri net allows for reliability modeling and reconfiguration, the performance analysis of the system is given via probabilistic model checking. The models are experimentally evaluated and compared. With the current widespread deployment of multi-core processors, one question in software engineering is how to effectively harness the parallel computing power provided by these processors. The architecture presented here allows us to explore the parallel computing power that otherwise may be wasted, and uses it to improve the dependability and survivability of a system, which is validated by our performance analysis.